Re: optimizer cost calculation problem

Tatsuo Ishii <t-ishii@sra.co.jp> writes:
> BTW it does not 2 gig, but 1 gig (remember that we do sortmembytes *
> 2) . 

Good point.  Probably that particular calculation should be
"sortmembytes * 2.0" to force it to double before it can overflow.
But I still think we'd better limit SortMem so that the basic
SortMem*1024 calculation can't overflow (or even come close to overflow,
likely).
        regards, tom lane